报告题目：Understanding Employee Ridesharing Incentives: Choice Modeling with Reinforcement Learning
Ridesharing has been a popular transportation means as it reduces travel costs, road congestion, and greenhouse gas emissions. Many governments and employers offer incentive programs to encourage commuter ridesharing. For the success of these programs, it is critical to incentivize new users’ adoption and retain existing users’ engagement by providing the best ridesharing experience possible. Using emerging information technology, such as mobile apps, it is now easy to fulfill ridesharing requests at any time. The ridesharing records collected via these mobile apps allow us to objectively learn employees’ usage patterns and quantify the risks relative to different transportation modes for the work commute. In particular, we propose an HEV-IRL model, which uses a choice model to account for variant risk levels among transportation modes and capture personalized utility. Its parameters are learned from longitudinal ridesharing records in the framework of reinforcement learning. This integrative model is empirically evaluated using a real-world employee ridesharing program dataset. We find that ridesharing imposes more risk than public transit or solo driving, and this risk is higher for passengers than for drivers. Though financial incentives are critical, social relationships with colleagues drive the adoption of the ridesharing service among company employees, especially drivers. As demonstrated by our simulation study, discerning users’ incentives can inform the design of the ridesharing matching system that will improve user experience and retention.